Of the 1024 total subjects enrolled in the Heart and Soul Study, we were able to measure ESV in the majority (989/1024, 97%). For the remaining 35 patients (3% of the cohort), ESVI could not be measured because of poor acoustic windows. Of the 989 subjects with ESVI data, 109 (11%) were hospitalized for HF during 3.6 ± 1.1 years (3485 person-years) of follow-up. The mean lengths of follow-up from the first to fourth quartiles of ESVI were 3.8 ± 0.9 years, 3.6 ± 0.9 years, 3.7 ± 1.1 years, and 3.2 ± 1.5 years, respectively ( P = 0001 by Kruskal-Wallis statistic). The highest quartile of ESVI had the shortest mean length of follow-up because more patients died in this subgroup. Compared with participants in the lower ESVI quartiles, those in the highest quartile were more likely to be male, to have a history of HF and MI, and to be receiving a renin-angiotensin system antagonist, beta-blocker, and diuretic therapy (). Those in the highest ESVI were also more likely to have a higher NT-proBNP level, a lower BMI, and lower systolic and diastolic blood pressures. Serum creatinine was similar between groups, but 24-hour urinary creatinine clearance was lower in participants in the highest ESVI quartile.
Baseline characteristics of participants by end-systolic volume index quartile
By echocardiography, participants in the highest ESVI quartile had a lower EF, higher EDVI, higher left atrial volume index, higher LV mass index, lower A-wave velocity, and a higher E/A ratio. Participants in the highest ESVI quartile were more likely to have diastolic dysfunction, mitral regurgitation of moderate or greater severity, and inducible ischemia (). Participants in the highest ESVI quartile had a mean ESVI of 37.5 ± 15.6 mL/m2. The median ESVI among participants in the highest quartile was 32 mL/m2, and the interquartile range was 26.7 to 41.8 mL/m2.
Echocardiographic characteristics of participants by end-systolic volume index quartile
During 3.6 ± 1.1 years of follow-up, we observed increased HF hospitalization and mortality among participants in the highest ESVI quartile (). In the unadjusted analysis, 67 (27%) participants in the highest ESVI quartile (ESVI > 25 mL/m2) were hospitalized with HF, compared with 8 (3%) of those participants in the lowest ESVI quartile (<13 mL/m2) (HR 9.7, 95% confidence interval [CI], 4.6–20.1; P < .0001). shows the unadjusted smoothed LOWESS plot relating ESVI to the proportion of participants hospitalized with HF. The association of ESVI with HF hospitalization was unchanged after adjustment for potential confounders and mediators (HR 5.0, 95% CI, 1.5–16.9; P = 01) (). compares the adjusted cumulative hazard of HF hospitalization for participants by ESVI quartile. In the above multivariable analyses, the referent group was the first quartile of ESVI. Using the first, second, and third quartiles as a combined referent group did not eliminate the association of ESVI > 25 mL/m2 with HF hospitalization (adjusted HR 3.1, 95% CI, 1.5–6.7; P = .003).
Adverse cardiovascular outcomes by ESVI quartile. *Combined end point = HF hospitalization or death.
LOWESS plot of the proportion of patients hospitalized with HF during follow-up according to baseline LV ESVI. LOWESS, Locally weighted, smoothed scatterplot.
Association of end-systolic volume index with heart failure hospitalization
Figure 3 Cumulative risk of HF hospitalization by quartile of ESVI. Curves were adjusted for sex, body mass index, history of HF, history of MI, medication use, measured creatinine clearance, log NT-proBNP, and echocardiographic variables (EF, LV EDVI, left atrial (more ...)
The highest quartile of ESVI predicted mortality on unadjusted analyses (HR 3.4, 95% CI, 2.1–5.6; P < .0001) but not on multivariate analysis (P = .70). The highest quartile of ESVI did, however, predict the combined end point of HF hospitalization or death on both unadjusted (HR 9.9, 95% CI, 4.8–20.6; P < .0001) and multivariable-adjusted analyses (HR 4.4, 95% CI, 1.3–15.4; P = .019). To examine the utility of ESVI as a predictor of incident HF hospitalization, we conducted an analysis of the 815 patients without a history of HF hospitalization. On unadjusted analysis, the highest quartile of ESVI predicted HF hospitalization (HR 9.7, 95% CI, 3.8–24.9; P < .0001). The association of ESVI with incident HF hospitalization persisted after adjustment for known cardiovascular risk factors, medication use, blood pressure, BNP, and renal function (adjusted HR 4.6, 95% CI, 1.3–16.1; P = .02). The highest quartile of ESVI did not predict incident HF hospitalization when other echocardiographic variables were included in multivariable Cox regression analysis.
To explore the association between ESVI and HF hospitalization among participants with preserved EF, we excluded all 109 participants with an EF less than 50%. Because participants in this subgroup had smaller LV volumes, we calculated new quartiles of ESVI that were specific to this subgroup (fourth quartile mean 27 ± 6 mL/m2, range 21–59 mL/m2). In those with EF ≥ 50%, even with a lower cutoff of ESVI for this subgroup, ESVI continued to predict increased HF hospitalization. Of those with an EF ≥ 50%, 37 participants (17%) in the highest ESVI quartile were hospitalized with HF, compared with 7 (3%) among those participants in the lowest ESVI quartile (unadjusted HR 5.4, 95% CI, 2.4–12.2; P < .0001). This association persisted after multivariable adjustment (adjusted HR 4.8, 95% CI, 1.0–23.2; P = .048). Of note, for all of the above multivariate analyses (for all outcomes and for all subgroup analyses), removing EDVI and EF to avoid multicollinearity had no effect on our results.
To examine the relative utility of ESVI as a diagnostic test to predict HF hospitalization, we analyzed ROC curves for LV end-systolic dimension index, ESVI, EDVI, and EF as predictors of this outcome. shows that the area under the ROC curve for ESVI was far superior to LV end-systolic dimension index (0.78 [95% CI, 0.71–0.85] vs 0.56 [95% CI, 0.47–0.65], respectively, P < .0001). The area under the ROC curve for ESVI was also superior to EDVI (0.71, 95% CI, 0.65–0.76; P < .0001) and EF (0.73, 95% CI, 0.68–0.78; P < .0001). Finally, we constructed a HR plot () to display the utility of an ESVI ≥ 25 mL/m2 as a predictor of HF hospitalization in various subgroups (HRs compare ESVI ≥ 25 mL/m2 vs ESVI < 25 mL/m2). This unadjusted analysis shows that ESVI predicts risk for HF hospitalization in a wide variety of patient subgroups.
ROC curves for LV ESDI, ESVI, and EF as predictors of hospitalization for HF. LV, Left ventricular.
Figure 5 HR plot for the risk of hospitalization for HF in ESVI ≥ 25 mL/m2 (vs ESVI < 25 mL/m2) in specified subgroups. *Cutoffs for NT-proBNP were determined by median BNP (173 pg/mL). †Only 4 patients with EF < 50% had an ESVI (more ...)